Description:FIELD OF THE INVENTION
The present disclosure relates to a method for predicting a likelihood of hypoxia. The present disclosure also relates to a system for predicting a likelihood of hypoxia
BACKGROUND OF THE INVENTION
Hypoxia is a condition where the body is deprived of oxygen and oxygen saturation level (SpO2) drops below 90%. Hypoxia can affect brain function, leading to problems with memory, concentration, and decision-making. Hypoxia also leads to tissue necrosis when prolonged. Moreover, prolonged hypoxia can damage vital organs like the heart, brain, and kidneys. Traditionally, hypoxia is screened with an oxygen saturation level measurement device which detects hypoxia only after oxygen saturation level has already dropped below certain value i.e., 90%. Furthermore, such device needs to be in contact with the subject for detecting hypoxia which may cause discomfort to the subject. The subject usually tends to remove the device as it feels uncomfortable adding challenge to continuous monitoring.
Therefore, in the light of foregoing discussion, there exists a need to overcome the aforementioned drawbacks.
SUMMARY OF THE INVENTION
A primary objective of the present disclosure seeks to provide a method for predicting a likelihood of hypoxia that provides a reliable and accurate prediction of hypoxia in a subject before actual occurrence of hypoxia using analysis of features extracted from micro-vibration based ballistocardiogram (BCG) like waveform data which can be obtained via non-contact sensors thereby eliminating discomfort of the subject. Moreover, such early prediction of hypoxia aids in providing care to prevent hypoxia from occurring rather than having to manage the condition after it occurs. Another objective of the present disclosure seeks to provide a system for predicting a likelihood of hypoxia using the aforementioned method. An aim of the present disclosure is to provide a solution that overcomes at least partially the problems encountered in prior art.
In a first aspect, an embodiment of the present disclosure provides a method for predicting a likelihood of hypoxia, the method comprising:
receiving a set of ballistocardiogram (BCG) waveform data;
deriving a set of signal dynamics data from the set of BCG waveform data;
applying at least one representation technique to the set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data;
applying a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data; and
predicting likelihood of hypoxia based on at least one dynamic feature.
In a second aspect, an embodiment of the present disclosure provides a system for predicting a likelihood hypoxia using method of claim 1, the system comprising a processor configured to:
receive a set of BCG waveform data;
derive a set of signal dynamics data from the set of BCG waveform data;
apply at least one representation technique to the set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data;
apply a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data; and
predict likelihood of hypoxia based on at least one dynamic feature.

The aforementioned method and system aids in integration of various waveform features (expressed in both frequency and time domain form) of various measured (via non-contact sensors) and derived vitals features such as heart rate, and respiration rate. These features, including those based on peak counting derived from the set of BCG waveform data are used to predict the likelihood of hypoxia. Therefore, comfort of subjects under monitoring is catered to which enhances patient experience during continuous and remote patient monitoring. Moreover, the aforementioned method and system utilize derivatives from BCG waveform data such as heart rate, respiration rate, spectral shift in power of BCG signal, morphology of BCG signal which diversify analysis of original BCG waveform data to provide an accurate and reliable relation between the BCG waveform data and hypoxia causing factors such as decreasing blood oxygen saturation level. Furthermore, the prediction can be made before actual onset of hypoxia which minimizes risk involved by alerting concerned medical practitioners and facilitates prompt provision of medical attention improving overall health outcomes. Furthermore, the aforementioned method and system integrate a machine learning model which accurately analyses the measured and derived vitals features of the set of BCG waveform data based on advanced analysis and prediction methodology (for example, time series analysis and recurrence analysis), for making such predictions for enhanced accuracy and minimizing human involvement thereby minimizing risk of human errors, therefore. Additionally, the machine learning model is configured to adapt to confounding factors that introduce noise by employing advanced signal processing techniques, feature selection methods, and adaptive filtering. Furthermore, the machine learning model is configured to utilizes real-time calibration and domain adaptation strategies to enhance robustness against environmental variations, and inconsistencies in data retrieval (such as erroneous data from malfunction sensor).
Embodiments of the present disclosure substantially eliminate or at least partially address the aforementioned problems in the prior art, and provide accurate prediction on likelihood of hypoxia occurrence in a subject, thereby minimizing the chance of/risk of potential harm and allowing for enhanced patient monitoring .
Additional aspects, advantages, features and objects of the present disclosure would be made apparent from the drawings and the detailed description of the illustrative embodiments construed in conjunction with the appended claims that follow.
It will be appreciated that features of the present disclosure are susceptible to being combined in various combinations without departing from the scope of the present disclosure as defined by the appended claims.
BRIEF DESCRIPTION OF THE DRAWINGS
The summary above, as well as the following detailed description of embodiments, is better understood when read in conjunction with the appended drawings. For the purpose of illustrating the present disclosure, exemplary constructions of the disclosure are shown in the drawings. However, the present disclosure is not limited to specific instrumentalities disclosed herein. Moreover, those skilled in the art will understand that the drawings are not to scale.
Embodiments of the present disclosure will now be described, by way of example only, with reference to the following diagrams wherein:
FIG. 1 illustrates a flow chart depicting steps of a method for predicting likelihood of hypoxia, in accordance with an embodiment of the present disclosure;
FIG. 2A illustrates a system for predicting likelihood of hypoxia, in accordance with an embodiment of the present disclosure;
FIG. 2B illustrates an environment implementing the system of FIG. 2A for predicting likelihood of hypoxia, in accordance with an embodiment of the present disclosure;
FIGs. 3A to 3E illustrate graphical representations of the vitals of five subjects in non-hypoxic state, in accordance with an embodiment of the present disclosure;
FIGs. 3F to 3J illustrate graphical representations of the vitals of five subjects in hypoxic state, in accordance with an embodiment of the present disclosure;
FIGs. 4A to 4E illustrate graphical representations of second harmonics analysis for a set of BCG waveform data corresponding to five subjects in non-hypoxic state, in accordance with an embodiment of the present disclosure;
FIGs. 4F to 4J illustrate graphical representations of second harmonics analysis for a set of BCG waveform data corresponding to five subjects in hypoxic state, in accordance with an embodiment of the present disclosure;
FIGs. 5A to 5E illustrate recurrence plots plotted for a set of signal dynamics data corresponding to five subjects in non-hypoxic state, in accordance with an embodiment of the present disclosure; and
FIGs. 5F to 5J illustrate recurrence plots plotted for a set of signal dynamics data corresponding to five subjects in hypoxic state, in accordance with an embodiment of the present disclosure.
In the accompanying drawings, an underlined number is employed to represent an item over which the underlined number is positioned or an item to which the underlined number is adjacent. A non-underlined number relates to an item identified by a line linking the non-underlined number to the item. When a number is non-underlined and accompanied by an associated arrow, the non-underlined number is used to identify a general item at which the arrow is pointing.
DETAILED DESCRIPTION OF EMBODIMENTS
The following detailed description illustrates embodiments of the present disclosure and ways in which they can be implemented. Although some modes of carrying out the present disclosure have been disclosed, those skilled in the art would recognize that other embodiments for carrying out or practising the present disclosure are also possible.
Referring to FIG. 1, illustrated is a method for predicting a likelihood of hypoxia. In this regard, throughout the present disclosure the term "hypoxia" refers to a state of medical emergency where tissues are deprived of adequate oxygen supply over a certain period of time. In this context, at step 102 a set of ballistocardiogram (BCG) waveform data is received. In this regard, the term "ballistocardiogram (BCG) waveform" refers to signal representing vibrations caused by motions of the body produced by various activity of the cardiovascular and respiratory system such as blood pumping via heart, inhalation-exhalation of air by lungs and so on.
In an embodiment the set of BCG waveform data pertains to at least one of: BCG signal, heart rate, respiration rate. In this regard, it may be appreciated that the BCG waveforms comprises waveforms related to both cardiac and respiratory activities. For example, various cardiac activity like blood pumping are indicated in the BCG waveforms. Notably, the BCG waveforms may be segregated using suitable filtering techniques to derive waveforms related to both cardiac and respiratory activities separately. Notably, waveforms related to respiratory activities (for sake of clarity named as respiratory component) typically have lower frequencies compared to the waveforms related to cardiac activities (for sake of clarity named as cardiac component). So, filtering BCG waveforms, for example using a filter, can help isolate the waveforms related to cardiac activities from the waveforms related to respiratory activities. It may be appreciated that the filter is at least one of: a frequency filter, a band pass filter, low pass filter, notch filter, digital filter such as infinite impulse response (IIR) filters and finite impulse response (FIR) filters and so on. In this regard, the waveforms related to respiratory activities extracted from the BCG waveform can be used to obtain data related to respiratory rate or to assess breathing patterns. Similarly, the waveforms related to cardiac activities derived can be used for obtaining data related to heart rate, blood flow volume and so on. Such data i.e., data obtained from both the waveforms related to cardiac activities and waveforms related to respiratory activities of the BCG waveforms, are referred to as the set of BCG waveform data.
Notably, the set of BCG waveform data are obtained from at least one sensor, preferably, at least one non-contact sensor which are used to monitor a subject. The at least one sensor may be piezo-electric sensors, pressure sensors, and/or any such suitable sensors that generates BCG signals based on vibration of the body. The term "subject" refers to a patient, a person, an animal, who require medical monitoring due to a medical condition, as part of a clinical assessment, as part of early screening in non-clinical subjects. The term "medical condition" refers to state which requires medical intervention such medications, and/or surgery. For example, the medical condition refers to heart disease (such as congestive heart failure, cardiac infection and the like), respiratory disease (such as SARS, pneumonia, influenza, chronic obstructive pulmonary disease (COPD), COVID-19, interstitial lung disease, pulmonary fibrosis, acute respiratory distress syndrome (ARDS), sepsis and the likes), respiratory failure, chronic respiratory conditions (such as asthma), decrease in organ performance, fresh transplant patient (heart, lungs, or any other organ), under immunosuppressant, prescribed with anaesthesia or sedative/analgesic medications, premature neonates in neonatal intensive care unit (NICU), in a comatose state, neurological disease/condition, obstructive sleep apnea and the likes.
Referring to FIG. 1, at step 104 a set of signal dynamics data is derived from the set of BCG waveform data. In this regard, after receiving BCG waveform data from the at least one sensor, the set of BCG waveform data (for example, the data pertaining to the heart rate and respiration rate) are analysed to derive the set of signal dynamics data. In this regard, the term "set of signal dynamics data" as used herein refers to data that changes over a time period. It may be appreciated that the set of signal dynamic data is related to both the cardiac component and the respiratory components of the BCG waveforms. In this regard, it may be appreciated that the set of BCG waveform data is collected over a time period at successive time intervals i.e., in a time series form. As the set of signal dynamics data is derived from the set of BCG waveform data, the set of signal dynamics data is also in time series form.
In an embodiment, the set of signal dynamic data pertains to at least one of: heart rate variability, respiration depth, inspiration/expiration rate, inspiration-expiration ratio, spectral shifts in power level of the BCG waveform data and morphology of the BCG waveform data. In this regard, heart rate variability (HRV) is a quantitative measure of the variation in time intervals between successive heartbeats and is calculated by analyzing the time differences between consecutive heartbeats from the waveforms related to cardiac activities of the BCG waveform data. Similarly, respiration depth, inspiration/expiration rate, inspiration-expiration ratio can be determined from the waveforms related to respiratory activities of the BCG waveform data. It may be appreciated that the spectral shifts in power level of the BCG waveform data refers to changes in the frequency distribution and amplitude of the BCG waveform data over time and can provide insights into cardiovascular and respiratory health. Similarly, morphology of the BCG waveform data refers to information on shape, amplitude, peaks and valleys of BCG waveforms. Notably, the set of signal dynamics data is derived using a suitable technique. For example, the set of signal dynamics data is derived using at least one of: standard deviation of normal-to-normal intervals (SDNN), root mean square of successive differences (RMSSD), frequency-domain metrics, Hilbert transform, empirical mode decomposition (EMD), wavelet transform (WT), using approximate entropy (ApEn) and sample entropy (SampEn). The technical advantage of deriving the set of signal dynamics data from the set of BCG waveform data is that the set of signal dynamics data provides rich, high-resolution insights into heart and lung function for early detection of likelihood of hypoxia.
With reference to FIG. 1, at step 106 at least one representation technique is applied to the set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data. In this regard, the term "at least one representation technique" refers to ways to represent the set of signal dynamics data (which are in time series form) in multi-dimension format. For example, phase space representation using time embedding can be used, where delayed versions of the signal are projected into a higher-dimensional space. In this regard, by using at least one representation technique, a vector may be constructed which is inclusive of multiple time steps from reaching one data value of the set of signal dynamics data to another. For instance, a set of signal dynamics data comprises values of HRV, such as v1 at time t1 and v2 at time t2. When at least one representation technique is applied, then a vector can be constructed which includes time steps (for example, t-12, t-24 are the time stamps in between t1 and t2) via which v1 is changed to v2, such as [v1, v(t-12), v(t-24), v2]. Notably, the term "time steps" refers to a predefined window of time over which measurement for the set of BCG waveform data is captured. Notably, when a machine learning model is implemented to construct the multidimensional time-embedded vector, then in such cases the machine learning model is configured for structuring past and present values together, over a time period defined on basis of available BCG waveform data. In other words, in such cases, instead of treating each time step as an independent observation, it preserves dependencies between consecutive time steps to provide a more accurate representation of temporal dynamics and underlying patterns in the set of BCG waveform data.
The term "multidimensional time-embedded vector" as used herein refers to a vector (for example, in a matrix from) including time steps that must be considered to determine how one value of the set of signal dynamics data changes to another value of the set of signal dynamics data. In other words, in the multidimensional time-embedded vector each data point is expressed as a vector consisting of values from multiple time steps. This approach captures state transitions by structuring past and present data values of the set of signal dynamics data, into a higher-dimensional space, preserving temporal dependencies within the signal dynamics data. . It may be appreciated that the multidimensional time-embedded vector can be used to capture temporal dependencies between values of the set of signal dynamics data, and corelate the set of signal dynamics data to blood oxygen saturation (SpO2) data to capture trend in how changes in the set of signal dynamics data is related to changes in blood oxygen saturation (SpO2) data. For example, the multidimensional time-embedded vector (constructed using at least one representation technique) can provide insight on if a HRV value is linearly increased/decreased to reach v2 from v1, or the HRV value is suddenly increased from v1 to an intermediate value (for example, v3) and then decreased to reach v2.
In an embodiment, the delayed version of the set of signal dynamics data comprises values of the set of signal dynamics data calculated at a predefined delay factor, using the at least one representation technique from amongst the set of representation technique. In this regard, the delayed version of the set of signal dynamics data refers to creating multiple shifted (or delayed) versions of the original the set of signal dynamics data (expressed in time-series format). Notably, the predefined delay factor and the dimension of the multidimensional time-embedded vector are determined based on volume of set of signal dynamics data. Notably, for a machine learning model implemented in order to predict likelihood of hypoxia, delay factor is a hyperparameter and can be tuned as per volume of set of signal dynamics data and other features that should be considered while making such prediction.
For example, the set of signal dynamics data in time-series format is expressed as: a(t), where 't' is the time, then the multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data (constructed using the at least one representation technique) is represented as: [a(t),a(t−τ),a(t−2τ),a(t−3τ)] where 'τ' is the predefined delay factor. For instance, if a dimension is set at 4 and the predefined delay factor is set at 12, then the multidimensional time-embedded vector would include: [a(t), a(t−12), a(t−24), a(t−36)]. Here, the constructed multidimensional time-embedded vector is represented with an embedding dimension of '4' and delay factor of '12'. It means that the original signal dynamics data is projected into '3' more dimensions to get a 4-dimensional space representation. Each projection (each column and row value) is a delayed version of original signal dynamics data, where the time delay is a multiple of 12. In this representation, each state does not just represent [a(t)] i.e., the value of the time series at time t. Instead, the constructed multidimensional time-embedded vector represents [a(t), a(t-12), a(t-24), a(t-36)]. So, for the states of the original signal dynamics data at time t1 and t2 to be considered similar, it is not enough for a(t1) and a(t2) to be comparable, their delayed versions also need to be comparable i.e. a(t1-12) ~ a(t2-12), a(t1-24) ~ a(t2-24) and a(t1-36) ~ a(t2-36) also needs to be true. This process (i.e., process of representing the set of signal dynamics data in form of multidimensional time-embedded vector) transforms the one-dimensional time-series data into a higher-dimensional space, enabling the analysis of temporal patterns and state transitions. The technical advantage is such multidimensional time-embedded vector (comprising delayed version of the set of signal dynamics data) enables a system to capture not just the current state of the set of signal dynamics data but also a path it has taken to reach that state for understanding temporal dependencies and trends in the set of signal dynamics data (which can be correlated to blood oxygen saturation (SpO2)) to predict likelihood of hypoxia.
In an embodiment, the set of representation technique comprises at least one of: a multi-dimensional delayed time series, delay embedding, dynamic mode decomposition, state-space representation, convolutional time delay networks, Fourier Transform, frequency domain representation, wavelet transform, autocorrelation, cross-correlation, recurrence plots, principal component analysis, Markov Model, Hidden Markov Model, Short-time Fourier Transform (STFT), Kalman filtering. In this regard, at least one of the aforementioned representation techniques can be used to trace a data value from another data value to construct the multidimensional time-embedded vector. For example, delay embedding creates delayed versions of the signal to capture temporal dependencies, while recurrence plots visualize state transitions, helping to identify patterns in the multidimensional representation. It may be appreciated that, if required, several of the aforementioned representation techniques can be combined to construct the multidimensional time-embedded vector. For instance, both multi-dimensional delayed time series and state-space representation may be used together to capture both temporal and spatial dynamics of the set of signal dynamics data. The technical advantage of these techniques is their ability to accurately transform the data values into higher-dimensional spaces, revealing hidden patterns and dynamics. For example, recurrence plots can reveal stable and unstable regions in the system, while frequency domain analysis can identify shifts in dominant frequencies, which may indicate changes in respiratory effort which can be linked to blood oxygen saturation (SpO2) to predict likelihood of hypoxia. Additionally, techniques like wavelet transform and Kalman filtering can reduce noise in the set of signal dynamics data (derived from the set of BCG waveform data), improving the accuracy of the analysis. By uncovering recurring patterns, trends, and anomalies in the data, these techniques enable a comprehensive understanding of the subject’s state/health status (specifically to blood oxygen saturation (SpO2)) to predict likelihood of hypoxia, which may not be apparent in the raw data values.
At step 108, a recurrence analysis is applied on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data. In this regard, when recurrence analysis is applied on the multidimensional time embedded vector to identify repetitive patterns of data values represented in vector form, over time. The recurrence analysis can be expressed in matrix wherein
Ri,j={█(1,&if ‖Xi-Xj ‖<€@0,&otherwise)┤
wherein € is a threshold for closeness which defines how close two points in the phase space must be to be considered recurrent. The threshold of closeness is determined empirically or based on statistical analysis of the set of signal dynamics data, ensuring that it captures meaningful recurrences without being overly sensitive to noise determined based on volume of the. The recurrence analysis can also be expressed in recurrence plot, comprising scattered points representing changes in data values over time. The term "recurring states" refers to instances where a data value (expressed in multidimensional vector form) in the multidimensional phase space is similar to a previously observed data value, indicating periodicity in the set of signal dynamics data corelated to a health status of the subject under observation. The term "dynamic feature" refers to measurable characteristic or parameter derived from the set of signal dynamics data that exhibit significant changes in their state over time, as identified through recurrence analysis. These changes can be correlated to physiological conditions, such as fluctuations in blood oxygen saturation (SpO2), to predict hypoxia. For example, the set of signal dynamics data may include HRV and the inspiration-expiration ratio. Using a suitable representation technique, a multidimensional time-embedded vector is constructed for HRV and inspiration-expiration ratio data. When the constructed multidimensional time-embedded vector is analysed, HRV data shows significant changes in its state, as indicated by increased instability in the recurrence plot, while the inspiration-expiration ratio remains relatively stable. This suggests that HRV is the at least one dynamic feature which can be correlated with the likelihood of hypoxia. Moreover, unlike static features (which remain relatively constant, such as mean J-peak amplitude, average I-J interval and so on), dynamic features evolve and aid in real-time monitoring and early detection of conditions like hypoxia. In other words, static features, which remain relatively constant, are considered to be baseline characteristics, whereas dynamic features change over time, enabling real-time monitoring and early detection of conditions like hypoxia.
In an embodiment, at least one dynamic feature is determined by comparing the set signal dynamics data with a historical set of signal dynamics data derived from a historical set of BCG waveform data, and corresponding historical blood oxygen saturation (SpO2) data, to corelate the set of signal dynamics data to blood oxygen saturation (SpO2) data. In this regard, the historical set of signal dynamics data refers to is a dataset derived from previously recorded BCG waveforms (i.e., historical set of BCG waveform data) and their corresponding blood oxygen saturation (SpO2) data. Notably, the historical set of BCG waveform data, the corresponding historical blood oxygen saturation (SpO2) data and the historical set of signal dynamics data may be stored and retrieved from a data repository. It may be appreciated that the historical set of signal dynamics data serves as a reference or baseline for comparison to identify how changes in the current signal dynamics data relate to changes in SpO2 levels, enabling the detection of conditions like hypoxia.
In this regard, by comparing the set of signal dynamics data namely, a present set of signal dynamics data with historical set of signal dynamics data, specifically, after identification of at least one dynamic feature, a correlation between present signal dynamics data and blood oxygen saturation (SpO2) levels can be established. The comparison is performed by analyzing patterns, trends, or anomalies in the present set of signal dynamics data and matching them with similar patterns observed in the historical set of signal dynamics data under specific SpO2 conditions. It may be appreciated that to ensure accurate comparison, noise in the present set of signal dynamics data is mitigated through suitable preprocessing techniques, such as smoothing or filtering. In other words, at least one dynamic feature is the determined as the parameter derived from the set of signal dynamics data exhibits significant change similar to a previous occurrence in the historical set of signal dynamics data under specific blood oxygen saturation (SpO2) condition. The technical advantage is accurate and precise identification of the at least one dynamic feature, such as HRV, for predicting likelihood of hypoxia. For instance, HRV data showed significant changes in its state during hypoxia events, as indicated by increased instability in the recurrence plot, enabling early detection of hypoxia.
In an embodiment, the at least one dynamic feature is indicative of likelihood of hypoxia when an increase in value of one or more signal dynamics data from amongst the set of signal dynamics data, compared to a predefined baseline value, is observed. In this regard, if value of any parameter from amongst the set of signal dynamics data, changes (for example, increases or decreases) beyond the predefined baseline value then it indicates that the subject may be at risk of onset of hypoxia. The predefined baseline value refers to nominal or standardized value for the parameter from amongst the set of signal dynamics data. For example, predefined baseline value of HRV measured using SDNN is 50-150ms. For another example, predefined baseline value for inspiration-expiration ratio is 1:1.5 to 1:2 and if the inspiration-expiration ratio for certain subject is 2:1, or 1:5 then such values indicate abnormality, and which may indicate laboured respiratory activity and possibly lead to hypoxia. The technical effect is accurate assessment and analysis of derived data to determine likelihood of hypoxia.
In an embodiment, the at least one dynamic feature is indicative of hypoxia onset comprises an increase in standard deviation values, coefficient of variation values and/or max-min difference calculated for the set of BCG waveform data, as compared to predefined nominal value. In this regard, the increase in standard deviation values and/or coefficient of variation values calculated for the set of BCG waveform indicates increased heart rate and respiration rate which suggest likelihood of hypoxia. For example, during hypoxia or before hypoxia, the heart rate increases by at least 15BPM and respiration rate increases by at least 5 respirations/min. By using standard deviation values and/or coefficient of variation values, noises, glitches which may be present in the heart rate and respiration data may be eliminated and a more reliable analysis on the set of BCG waveform may be performed. Thus, accurate identification of at least one dynamic feature can be made for reliable predication of likelihood of hypoxia. Similarly, from the difference calculated for the set of BCG waveform data (specifically, for the heart rate and respiration rate data) as compared to predefined nominal value provides insight into how the set of signal dynamics data are changing. The observed changes can be correlated to blood oxygen saturation (SpO2) change to predict likelihood of hypoxia.
It may be appreciated that changes in the heart rate, changes in respiration rate (indicated by breathing pattern), irregular rhythms of the heart or breathing may suggest respiratory distress and can also signal potential hypoxia. To quantify these changes in terms of to identify at least one dynamic feature, suitable statistical technique may be applied such as calculation of standard deviation and coefficient of variation of received set of BCG waveform data, specifically, respiration rate and heart rate. Therefore, a relation between the set of BCG waveform data (i.e., heart rate and respiration rate) and likelihood of hypoxia can be established. For example, the standard deviation and/or coefficient of variation of either heart rate or respiration rate or both during the hypoxia tends to be higher. Therefore, if an incremental/upward trend in values of the standard deviation and/or coefficient of variation is observed from the set of BCG waveform data then a possible onset of hypoxia may be predicted. It may be appreciated that changes in at least one of orientation, propagation, or trend of BCG signal/waveform characteristics in relation to physiological events (such as change in respiratory or cardiac activities) can also be corelated to likelihood of hypoxia. Notably, considering the changes in heart rate, respiration rate, orientation, propagation, or trend of BCG signal/waveform characteristics aids in minimizing noises and errors due to artifact rejection (e.g., distinguishing physiological trends from sensor drift) while improving personalized monitoring by adapting machine learning models to individual waveform variations to provide accurate and early detection of cardiovascular conditions by tracking progressive changes and identifying deviations from normal function.
At step 110, likelihood of hypoxia is predicted based on the at least one dynamic feature. In this regard, when the at least one dynamic feature is identified by applying recurrence analysis on the set of signal dynamics data, such identified dynamic feature is associated with hypoxia, for example, increased HRV value, increased inspiration-expiration ratio or so on are indicative of increased labour on cardiopulmonary system (heart and lungs) of the subject which is related to oxygen demand of body and can suggest that subject is lacking in required oxygen and can go into hypoxia if this continues. The technical advantage is predicting hypoxia before actual happening and thus provisions can be made to avoid hypoxia thereby providing better health care to the subject.
In this regard, it may be appreciated that from the BCG waveform data (that comprises BCG signal, hear rate and respiration rate data), the heart rate data and the respiration rate data is extracted. The data is expressed in multidimensional time-embedded vector from which trends in heart rate and respiration rate data values are identified. Then from the trends of heart rate and respiration rate data values, at least one dynamics features (which is a metric or parameter) is identified. At least one dynamic feature (calculated as standard deviation values, coefficient of variation values and/or max-min difference calculated for the set of BCG waveform data) can be used to distinguish hypoxic state of the subject (i.e., hypoxic region in a recurrence plot indicating regions with low SpO2) from non-hypoxic regions (i.e., non-hypoxic region in a recurrence plot indicating regions with normal SpO2). Moreover, by tuning the length of a filter window, the trends can also be analysed to identify the at least one dynamics features which can be co-related to likelihood of hypoxia. It may be appreciated that the filter window comprises a baseline window and a comparison window. Notably, using multiple filter windows aids in accurately predicting the possible onset of hypoxia in advance. For example, the trends in heart rate and respiration rate is analysed by comparing in heart rate and respiration rate data values over time, by using a 30-minute baseline window and a 15-minute comparison window, a median heart rate change and median respiration rate change are tracked at 15-minute intervals. If sufficient valid data points exist within this time frame, both percentage and absolute changes in heart rate and respiration rate is computed, in order to ascertain the trends and assess fluctuations in vitals over time.
Furthermore, it may be appreciated that the likelihood of hypoxia can also be predicted from analysis of the morphology and spectral shift in power of the BCG signal. In this regard, the recurrence analysis can be applied thereto, and recurrence plots can be generated. For an example, recurrence plot can be generated for second harmonic power of the BCG signal (which comprises cardiac component). Notably, the BCG signal is first filtered through signal processing module to extract the cardiac component thereof. Then the power of the second harmonic frequency of that cardiac component is computed as a continuous time series (using a suitable representation technique). Recurrence plots are generated through phase space representations and time embedding of the time series representation of the cardiac component. Notably, changes in the recurrence plots start corresponding to change in the BCG signal is observed as the subject’s health status changes i.e., when the subject goes from a non-hypoxic to hypoxic state. Notably, recurrence plot refers to a visual representation of a time series representation of the set of signal dynamics data, that shows when the state of the set of signal dynamics data returns to a previously visited state, basically, repeated patterns in chaotic state of signal dynamics data. It may be appreciated that recurrence plot refers to a phase space representation using time embedding where delayed versions of the same signal are projected into multidimensional space for a multidimensional representation of the behaviour or trend of the set of signal dynamics data.
In an embodiment, the aforementioned method further comprising step of integrating a machine learning model for predicting onset of hypoxia, wherein the machine learning model is configured to:
receive a historical set of BCG waveform data;
derive a historical set of signal dynamics data from the historical set of BCG waveform data;
applying at least one representation technique to the set of signal dynamics data and the historical set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data and the historical set of signal dynamics data;
applying a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data; and
predicting onset of hypoxia based on the at least one dynamic feature.
The machine learning (ML) model is configured to execute various steps to predict likelihood of hypoxia. In this regard, the ML model is configured to receive the historical set of BCG waveform data. The ML model also receives the corresponding historical blood oxygen saturation (SpO2) data. It may be appreciated that the ML model is configured to receive the historical set of BCG waveform data from the data repository when required. The ML model is configured to derive the historical set of signal dynamics data from the historical set of BCG waveform data by using the suitable technique. Notably, the ML model is also configured to deduce a relationship between the historical set of signal dynamics data and the historical blood oxygen saturation (SpO2) data. When the ML model is fed with the set of BCG waveform data, for sake of clarity termed as the present set of BCG waveform data, the ML model is configured to apply at least one representation technique to the set of signal dynamics data (derived from the present set of BCG waveform data) and to the historical set of signal dynamics data (derived from the historical set of BCG waveform data) for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data and the historical set of signal dynamics data. The ML model applies a recurrence analysis and/or any such suitable techniques (for instance, image processing techniques) on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data and predicts likelihood of hypoxia based on the at least one dynamic feature. Notably, such suitable technique may be at least one of nonlinear time series Analysis Lyapunov exponent estimation, functional connectivity analysis, Hidden Markov Models (HMMs), Poincaré maps and so on.
In an embodiment, the machine learning model is further configured to assign a risk score to a given dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data. The risk score may be expressed in terms of numeral scale, for example, in a scale of 0 to 5, to indicate in how much risk the subject is and what is the likelihood of hypoxia for the subject. For example, if the ML model assigns a risk score of 1 to a subject, then that subject may not need immediate medical attention for hypoxia, whereas is the assigned risk score is 3 or more then the subject requires immediate medical attention for hypoxia which may be happening soon. The technical advantage is implementing a systematic approach to generate alert, and to ease the approach of providing required medical attention to the subject.
In an embodiment, the machine learning model is configured to be fine-tuned if required based on new data (i.e., new set of BCG waveform data, and/or new set of signal dynamics data), to enhance accuracy of prediction based on changes in the data over time. The technical advantage is enhanced accuracy of prediction of hypoxia and enhanced performance of the ML model.
In an embodiment, the aforementioned method further comprises step of generating an alert indicative of the likelihood of hypoxia, wherein the alert is at least one of: a visual alert, a haptic alert, an audio alert. In this regard, when the likelihood of hypoxia is predicted based on the at least one dynamic feature, then the alert is generated. The alert may be a blink of light, flashing in a monitor, a vibration on a device such as a pager, mobile phone, smart phone, or smart wristwatch, a ringing or siren, a text on the monitor, the mobile phone, smart phone, or smart wristwatch or and so on. It may be appreciated that the ML model is also configured to generate the alert indicative of the likelihood of hypoxia.
With reference to FIG. 2A, illustrated is a system 200 for predicting likelihood of hypoxia using aforementioned method, in accordance with an embodiment of the present disclosure. As shown, the system 200 comprises a processor 202. The processor 202 may be a microcontroller, microprocessor, an on-chip control unit, a central processing unit, or any such suitable arrangement capable of receiving input, analyze the input received, and perform required operation thereafter.
It may be appreciated that various embodiments and variants disclosed above, with respect to the aforementioned method, apply mutatis mutandis to the system 200 as well.
The processor 202 is configured to receive a set of ballistocardiogram (BCG) waveform data. Notably, the BCG waveform data is captured by at least one sensor 204 communicably coupled to the processor 202. It may be appreciated that the processor 202 is also communicably coupled to a data repository 206 for receiving additional input, for example, the processor 202 is configured to receive processed data (i.e., historical set of BCG waveform data, corresponding blood oxygen saturation data, and historical set of signal dynamics data from the data repository 206. Moreover, the processor 202 is also configured to derive a set of signal dynamics data from the set of BCG waveform data received and to apply the at least one representation technique to the set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data. The processor 202 is also configured to apply a recurrence analysis on the multidimensional time-embedded vector constructed from the set of signal dynamics data for identifying recurring states and determining at least one dynamic feature from there amongst. Based on the identified at least one dynamic feature, the processor 202 is configured to predict likelihood of hypoxia. As shown, the processor 202 is also communicably coupled to a ML model 208 to predict likelihood of hypoxia.
In this regard, the processor 202 is configured to execute the steps of the aforementioned method to for predicting likelihood of hypoxia.
In an embodiment, the set of BCG waveform data pertains to at least one of: BCG signal, heart rate, respiration rate.
In an embodiment, the set of signal dynamics data pertains to at least one of: heart rate variability, respiration depth, inspiration/expiration rate, inspiration-expiration ratio, spectral shifts in power level of the BCG waveform data and morphology of the BCG waveform data.
In an embodiment, the delayed version of the set of signal dynamics data comprises values of the set of signal dynamics data calculated at a predefined delay factor, using the at least one representation technique from amongst the set of representation technique.
In an embodiment, the set of representation technique comprises at least one of: a multi-dimensional delayed time series, delay embedding, dynamic mode decomposition, state-space representation, convolutional time delay networks, Fourier Transform, frequency domain representation, wavelet transform, autocorrelation, cross-correlation, recurrence plots, principal component analysis, Markov Model, Hidden Markov Model, Short-time Fourier Transform (STFT), Kalman filtering.
In an embodiment, the processor is further configured to integrate a machine learning model for predicting onset of hypoxia, wherein the machine learning model is configured to:
receive a historical set of ballistocardiogram (BCG) waveform data;
derive a historical set of signal dynamics data from the historical set of BCG waveform data;
apply at least one representation technique to the set of signal dynamics data and the historical set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data and the historical set of signal dynamics data;
apply a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data; and
predict onset of hypoxia based on the at least one dynamic feature.
In an embodiment, the machine learning model 208 is further configured to assign a risk score to a given dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data.
In an embodiment, the processor is configured to finetune the machine learning model if required based on new data (i.e., new set of BCG waveform data, and/or new set of signal dynamics data), to enhance accuracy of prediction based on changes in the data over time. The technical advantage is enhanced accuracy of prediction of hypoxia and enhanced performance of the ML model.
In an embodiment, the processor is further configured to generate an alert indicative of the onset of hypoxia, wherein the alert is at least one of: a visual alert, a haptic alert, an audio alert.
Referring to FIG. 2B, illustrated is an environment 210 implementing the system 200 of FIG. 2A for predicting likelihood of hypoxia, in accordance with an embodiment of the present disclosure. As shown, the system 200 monitors a subject 212 and obtains the set of BCG waveform data from the at least one sensor 204. The system 200 processes the received set of BCG waveform data to predict the likelihood of hypoxia in the subject. The environment 210 comprises an alert module 214 communicably connected to the system 200. The system 200 is configured to operate the alert module 214 based on prediction of likelihood of hypoxia.
FIGs. 3A to 3E illustrate graphical representations of vitals of five subjects in non-hypoxic state, in accordance with an embodiment of the present disclosure. As shown, the vitals recorded are heart rate, respiration rate and blood oxygen saturation (SpO2). In this regard, the vitals (i.e., heart rate, respiration rate and blood oxygen saturation (SpO2)) are recorded consistently over time. Notably, the values for the vitals are shown in Y-axis and the time of recording are shown in X-axis. As shown in FIGs. 3A to 3E, the graphs for the heart rate, and the respiration rate show no sudden increase of values. Similarly, the graphs for blood oxygen saturation (SpO2) demonstrate no sudden decrease of values. Thus, the concerned subjects are in a normal i.e., non-hypoxic state.
FIGs. 3F to 3J illustrate graphical representations of vitals of five subjects in hypoxic state, in accordance with an embodiment of the present disclosure. With respect to FIGs. 3F to 3J, the vitals recorded are heart rate, respiration rate and blood oxygen saturation (SpO2). The vitals are recorded consistently over time. Notably, the values for the vitals are shown in Y-axis and the time of recording are shown in X-axis. As shown, the graphs for the heart rate, and the respiration rate show sudden increase of values at certain time t1. Similarly, the graphs for blood oxygen saturation (SpO2) demonstrated sudden decrease of values correspondingly, at time t1, when the heart rates and the respiration rates show sudden increase. Thus, it can be asserted that when blood oxygen saturation (SpO2) decreases the heart rate and respiration rate increases. Such response indicates that the subjects are under deprivation of required oxygen and may go into hypoxic state in near future.
FIGs. 4A to 4E illustrate graphical representations of second harmonics analysis for a set of BCG waveform data corresponding to five subjects in non-hypoxic state, in accordance with an embodiment of the present disclosure. As shown, the fluctuations in the second harmonic power of the set of BCG waveform data namely, heart rate, for five subjects. In this regard, the X-axis represents time, and Y-axis represents power value of second harmonics of heart rate. Notably, the graphical representations show no sudden fluctuation in power value with change in time. Therefore, it can be concluded that the five subjects are in non-hypoxic state.
FIGs. 4F to 4J illustrate graphical representations of second harmonics analysis for a set of BCG waveform data corresponding to five subjects in hypoxic state, in accordance with an embodiment of the present disclosure. As shown, the graphical representations show sudden fluctuation in power value with change in time. Therefore, it can be concluded that the five subjects are in hypoxic state.
Referring to FIGs. 5A to 5E, illustrated are recurrence plots plotted for a set of signal dynamics data corresponding to five subjects in non-hypoxic state, in accordance with an embodiment of the present disclosure. In this regard, the recurrence plots are used to identify parameters from amongst the set signal dynamics data to quantify results from recurrence analysis applied on multidimensional time-embedded vector (constructed from the set signal dynamics data using at least one representation technique). Notably, from the recurrence plots provides insight on time steps at which the subject’s previous state recurs. Notably, the recurrence plots shown in FIGs. 5A to 5E are plotted for change in spectral shift in power level of the BCG waveform (i.e., the second harmonics analysis of heart rate plotted in FIGs. 4A to 4E). It may be appreciated that the recurrence plots can be plotted from any parameter from amongst the set of signal dynamics data. As shown, grey regions in the recurrence plots indicate stable region and black regions in the recurrence plots indicate unstable regions indicating hypoxia onset. In FIGs. 5A to 5E, the grey regions are widely observed in the recurrence plots. So, it can be concluded that the subjects are in non-hypoxic state.
FIGs. 5F to 5J illustrate recurrence plots plotted for a set of signal dynamics data corresponding to five subjects in hypoxic state, in accordance with an embodiment of the present disclosure. Notably, the recurrence plots shown in FIGs. 5F to 5J are plotted for change in spectral shift in power level of the BCG waveform (i.e., the second harmonics analysis of heart rate plotted in FIGs. 4F to 4J). Notably, the hypoxia can be clearly distinguished by looking at a relative fractional area of the grey region with respect to the black region. It may be appreciated that when the grey region is less than 0.75 of total plot area, then it can be concluded that the concerned subjects are in hypoxia state. As shown, the black regions are widely observed in the recurrence plots. So, it can be concluded that the subjects are in hypoxic state.
Modifications to embodiments of the invention described in the foregoing are possible without departing from the scope of the invention as defined by the accompanying claims. Expressions such as “including”, “comprising”, “incorporating”, “consisting of”, “have”, “is” used to describe and claim the present invention are intended to be construed in a non-exclusive manner, namely allowing for items, components or elements not explicitly described also to be present. Reference to the singular is also to be construed to relate to the plural. Numerals included within parentheses in the accompanying claims are intended to assist understanding of the claims and should not be construed in any way to limit subject matter claimed by these claims.
EXPERIMENTAL PART
An experiment on BCG waveform data collected for five subjects was conducted. The collected BCG waveform data was filtered, and cardiac component (waveforms related to cardiac activity) and respiratory component (waveform related to respiratory activities) were extracted. Such extracted data primarily pertains to heart rate and respiratory rate respectively.
The standard deviation, coefficient of variation of heart rate (HR), and max – min HR value during hypoxia and non-hypoxia state was computed and noted in tabulation 1 given below.
Tabulation 1
Subject ID standard deviation: non-hypoxia region standard deviation: hypoxia region coefficient of variation: non-hypoxia region coefficient of variation: hypoxia region max – min HR: non-hypoxia region max – min HR: hypoxia region
DZ_001 3.14 17.29 0.04 0.16 12.19 51.13
DZ_002 1.58 7.95 0.02 0.10 8.72 22.14
DZ_003 2.06 7.76 0.03 0.09 7.95 23.67
DZ_004 1.40 6.95 0.01 0.07 6.61 16.89
DZ_005 1.37 10.10 0.03 0.13 6.03 34.88
From the tabulation 1, it was observed that during hypoxia state, the standard deviation, the coefficient of variation, and max-min values of heart rate was increased. Thus, it was concluded that increases in the standard deviation, the coefficient of variation, and max-min values of heart rate was indicative on likelihood of hypoxia.
Similarly, the standard deviation, coefficient of variation of respiration rate (RR) and max – min RR value during hypoxia and non-hypoxia state was computed and noted in tabulation 2 given below.
Tabulation 2
Subject ID Standard Deviation: non-hypoxia region Standard Deviation: hypoxia region coefficient of variation: non-hypoxia region coefficient of variation: hypoxia region max – min RR: non-hypoxia region max – min RR: hypoxia region
DZ_001 1.72 1.58 0.08 0.05 5.88 6.90
DZ_002 0.91 3.88 0.03 0.13 3.13 12.03
DZ_003 0.87 1.87 0.06 0.09 3.53 7.41
DZ_004 0.78 2.51 0.03 0.08 3.90 10.43
DZ_005 0.95 2.17 0.05 0.09 4.33 8.36
From the tabulation 2, it was observed that during hypoxia state, the standard deviation, the coefficient of variation, and max-min values of respiration rate was generally increased. Thus, it was concluded that increases in the standard deviation, the coefficient of variation, and max-min values of respiration rate was indicative of likelihood of hypoxia.
, Claims:I/We claim:
1. A method for predicting a likelihood of hypoxia, the method comprising:
receiving a set of ballistocardiogram (BCG) waveform data;
deriving a set of signal dynamics data from the set of BCG waveform data;
applying at least one representation technique to the set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data;
applying a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data; and
predicting likelihood of hypoxia based on the at least one dynamic feature.
2. The method as claimed in claim 1, wherein the set of BCG waveform data pertains to at least one of: BCG signal, heart rate, respiration rate.
3. The method as claimed in claim 1, wherein the set of signal dynamics data pertains to at least one of: heart rate variability, respiration depth, inspiration/expiration rate, inspiration-expiration ratio, spectral shifts in power level of the BCG waveform data and morphology of the BCG waveform data.
4. The method as claimed in claim 1, wherein the delayed version of the set of signal dynamics data comprises values of the set of signal dynamics data calculated at a predefined delay factor, using a given representation technique from amongst the at least one representation technique.
5. The method as claimed in claim 1, wherein the representation technique comprises at least one of: a multi-dimensional delayed time series, delay embedding, dynamic mode decomposition, state-space representation, convolutional time delay networks, Fourier Transform, frequency domain representation, wavelet transform, autocorrelation, cross-correlation, recurrence plots, principal component analysis, Markov Model, Hidden Markov Model, Short-time Fourier Transform (STFT), Kalman filtering.
6. The method as claimed in claim 1, further comprising step of integrating a machine learning model for predicting onset of hypoxia, wherein the machine learning model is configured to:
receive a historical set of BCG waveform data;
derive a historical set of signal dynamics data from the historical set of BCG waveform data;
applying at least one representation technique to the set of signal dynamics data and the historical set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data and the historical set of signal dynamics data;
applying a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data; and
predicting onset of hypoxia based on the at least one dynamic feature.
7. The method as claimed in claim 6, wherein the machine learning model is further configured to assign a risk score to a given dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data.
8. The method as claimed in claim 1, wherein the at least one dynamic feature is determined by comparing the set signal dynamics data with a historical set of signal dynamics data derived from a historical set of BCG waveform data, and corresponding historical blood oxygen saturation (SpO2) data, to corelate the set of signal dynamics data to blood oxygen saturation (SpO2) data.
9. The method as claimed in claim 1, wherein the at least one dynamic feature is indicative of likelihood of hypoxia when an increase in value of one or more signal dynamics data from amongst the set of signal dynamics data, compared to a predefined baseline value, is observed.
10. The method as claimed in claim 9, wherein the at least one dynamic feature indicative of hypoxia onset comprises an increase in standard deviation values, coefficient of variation values and/or max-min difference calculated for the set of BCG waveform data, as compared to predefined nominal value.
11. The method as claimed in claim 1, further comprising generating an alert indicative of the onset of hypoxia, wherein the alert is at least one of: a visual alert, a haptic alert, an audio alert.
12. A system (200) for predicting a likelihood hypoxia using method of claim 1, the system comprising a processor (202) configured to:
receive a set of BCG waveform data;
derive a set of signal dynamics data from the set of BCG waveform data;
apply at least one representation technique to the set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data;
apply a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data; and
predict onset of hypoxia based on the at least one dynamic feature.
13. The system (200) as claimed in claim 12, wherein the processor (202) is further configured to integrate a machine learning model (208) for predicting onset of hypoxia, wherein the machine learning model is configured to:
receive a historical set of BCG waveform data;
derive a historical set of signal dynamics data from the historical set of BCG waveform data;
apply at least one representation technique to the set of signal dynamics data and the historical set of signal dynamics data, for generating a multidimensional time-embedded vector comprising delayed version of the set of signal dynamics data and the historical set of signal dynamics data;
apply a recurrence analysis on the multidimensional time-embedded vector for identifying recurring states and determining at least one dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data; and
predict onset of hypoxia based on the at least one dynamic feature.
14. The system (200) as claimed in claim 13, wherein the machine learning model (208) is further configured to assign a risk score to a given dynamic feature from amongst the set of signal dynamics data and the historical set of signal dynamics data.
15. The system (200) as claimed in claim 12, wherein the processor (202) is further configured to generate an alert indicative of the onset of hypoxia, wherein the alert is at least one of: a visual alert, a haptic alert, an audio alert.